Which are the long-run determinants of US outward FDI? Evidence using large long-memory panels

Abstract

This paper analyzes the long-run determinants of US outward FDI (OFDI) stock, focusing mainly on the Euro Area (EA) for the period 1985–2019. We consider a sample of 54 developed and emerging host countries representing over 70% of the total US OFDI stock. We aim to capture different determinants by country groups zooming in on the European Union (EU). We implement a Dynamic Common Correlated Effects Pooled Mean Group (DCCEPMG) estimator for this aim. Our econometric approach is especially suited for analyzing integrated economic areas as it allows us to deal with cross-section dependence (CSD), non-stationarity, structural breaks, and slope homogeneity usually present in large panel data. Our main results suggest that horizontal (HFDI) and vertical (VFDI) strategies coexist for all country groups. However, as we move towards more homogeneous groups, the results show the greater importance of VFDI. Additionally, we find that some variables have a common long-run effect on US OFDI, especially for smaller and more homogeneous groups.

Keywords:

• BMA
• DCCEPMG
• long-run
• structural breaks
• economic integration

JEL classifications:

• F21
• F23
• R39

Acknowledgments

This paper has benefited from the comments and suggestions received in presentations at the Economics Department of the University of Goettingen, the XVIII INTECO Workshop and the 24th INFER Annual Conference in Timisoara. The European Commission's support does not constitute an endorsement of the contents, which reflect the views only of the authors, and the Commission cannot be held responsible for any use that may be made of the information contained therein. All remaining errors are ours.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See Straathof et al. (2008), Bruno et al. (2017), Carril-Caccia and Pavlova (2018) and Bruno, Campos, and Estrin (2021), among others.

2 See the knowledge capital model of Carr, Markusen, and Maskus (2001).

3 The expansion and complexity of the production fragmentation across borders via Global Value Chains (GVCs) has led (Yeaple 2003) to coin the FDI generated by these mixed motives as ‘complex FDI’ and more recently (Baldwin and Okubo 2014) have developed the concepts of ‘horizontal-ness’ and ‘vertical-ness’ to systematically account for these more complex forms of FDI.

4 See Petroulas (2007), Brouwer, Paap, and Viaene (2008), Baldwin et al. (2008), De Sousa and Lochard (2011), Neary (2009) and Carril-Caccia and Pavlova (2018), Sondermann and Vansteenkiste (2019),among others.

5 We express our gratitude to a reviewer for highlighting this point. It is true that alterations in the US trade and security policies can play a significant role in influencing long-term US FDI decisions, and our study could capture these effects. The establishment of free trade agreements (FTAs), bilateral investment treaties (BITs), or even military alliances can potentially serve as drivers of FDI. However, a comprehensive investigation of this issue is outside the scope of this study and will be addressed in future research.

6 Most FDI has accumulated in the EU and in the North American Free Trade Agreement (NAFTA), each of them representing percentages close to 30%. The other main recipients are the Association of Southeast Asian Nations (ASEAN) and Japan, China, and the Republic of Korea (ASEAN plus Three), and Mercado Común del Sur (MERCOSUR).

7 This division is based not only on geographical criteria but also on economic similarities. Indeed, Bayoumi and Eichengreen (1993), Zhang and Artis (2001) and Konstantakopoulou and Tsionas (2011), among others, found that this classification could be based on business cycles synchronicity and common economic shocks.

8 See, among others, Pesaran (2006), Eberhardt and Bond (2009), Chudik and Pesaran (2015), and Ditzen (2018).

9 The US has signed FTAs with countries from different regions of the world. According to Villarreal and Fergusson (2014) and MacDermott (2007), the NAFTA has significantly increased US investment to Canada and Mexico. In the same vein, the Dominican Republic-Central America United States Free Trade Agreement (CAFTA-DR) has also stimulated US FDI flows to this region (Hornbeck 2012). US trade agreements may have encouraged VFDI to the detriment of HFDI. Osnago, Rocha, and Ruta (2019), who study how deep FTAs affect FDI originating in Japan, Germany and the US, show that the depth of trade agreements is correlated with VFDI. On the other hand, Im (2016) finds that FTAs where the US takes part have discouraged US HFDI. In addition, the US has also established several BITs with different parties. Haftel (2010) demonstrates that investment treaties have had a positive impact on US OFDI.

10 Apart from trade and investment agreements, military alliances and interstate security relations create also a favorable environment for the location of US FDI. Indeed, geopolitical stability could also contribute to the establishment of trade agreements (Eichengreen, Mehl, and Chitu 2019). Security factors are especially relevant in developing countries, as these regions tend to have higher political instability. According to Biglaiser and DeRouen Jr (2007), diplomatic factors are important determinants for US OFDI flows in developing countries. Similarly, Li and Vashchilko (2010) find that security alliances influence FDI flows from high income to low income countries. Against this background, the enlargements of the NATO towards Central and Eastern members of the EU may have also changed the drivers of US OFDI. Schweickert et al. (2011) show that pre-accession incentives provided by EU and NATO have influenced institutional development of these transition countries positively.

11 For example, Phillips and Moon (1999) and Kao (1999) establish the asymptotic normality of the within estimator for the cases in which regressors follow unit root processes.

12 This is the estimator that we use for the estimation of the long-run determinants of US OFDI stock.

13 See Lee and Pesaran (1993), Conley and Topa (2002), Conley and Dupor (2003), Pesaran, Schuermann, and Weiner (2004) and Dées et al. (2007), among others.

14 When the cross-section dimension is short, and the time-series dimension is long, the standard approach to dependence is to treat the equations from the different cross-section units as a system of seemingly unrelated regression equations (SURE) and then estimate the system by GLS techniques (See Ahn, Lee, and Schmidt 2001; Holtz-Eakin, Newey, and Rosen 1988; Kiefer 1980; Lee 1991). Nevertheless, in the first case, a distance measure is not always available, while the SURE-GLS approach involves nuisance parameters as the cross-section dimension of the panel increases (and becomes non-feasible when N>T). Moreover, the SURE estimator would not be consistent if the source of CSD is correlated with the regressors.

15 See Robertson and Symons (2000), Coakley, Fuertes, and Smith (2002) and Phillips and Sul (2003).

16 In contrast to Kao and Chiang (2000), Banerjee and Carrion-i Silvestre (2004), Westerlund (2006) or Gutierrez (2010), that assumed independence across units.

17 The Middle East and North Africa.

18 Brazil, Russia, India, China, and South Africa.

19 Poland, the Czech Republic, Slovakia, and Hungary.

20 We use stock data instead of flows because they are more persistent and reliable along time. Therefore, it is a better measure to study the long-run FDI determinants.

21 Note that some of the variables selected as robust in the BMA analysis cannot be directly translated into the cointegration analysis. These are notably dummies, that will be indirectly accumulated in the country fixed effects or captured by the structural breaks.

22 Moreover, using several variables that capture the same effect would generate multicolinearity in the empirical model. We are also limited by the degrees of freedom, so that we choose one representative (robust) variable from the different categories described in Table 1.

23 It could be due to spatial dependence, omitted unobserved common components, or idiosyncractic pair-wise dependence of $u_{i,t}$ and $u_{j,t}$ $(i\ne j)$ with no particular pattern of spatial or common components.

24 In a panel, as we are interested in obtaining the estimation of the long-run relationship before and after the break, we have to impose a single common break.

25 The list of variables and abbreviations can be found in Table 2.

26 Evidently, the larger is the number of cross-section units, the higher the potential degree of heterogeity. Pesaran et al. (1999) mention that, in the case of cross-country studies, the likelihood ratio test usually reject the hypothesis of equal error variances and/or slopes (short-run or long-run) at conventional significance levels.

27 For more information about competition forces and FDI location, see Crozet, Mayer, and Mucchielli (2004).

28 In particular, one unit increase in the tariff reduces US OFDI by approximately 85%. The possible explanation for this large effect is that revenue trade taxes is meager (between 0.5% and 1.5%) for the fundamentally open EU countries. Therefore, a 1 percentage point increase of this variable implies a doubling of the tariff.

29 In this case, we find a break in the mean and the trend of the relationship, but the cointegration vector is stable during the sample period.

Additional information

Funding

The authors acknowledge the financial support from the AEI-Spanish Ministry of Economy and Competitiveness (Ministerio de Economía y Competitividad) (Project number PID2020-114646RB-C42 / AEI 10.13039 / 501100011033). Cecilio Tamarit and Mariam Camarero also acknowledge the funding from the European Commission (Project numbers ERASMUS-JMO-2021-CHAIR 101047088 and ERASMUS-JMO-2022-CHAIR 101083430), respectively.